Principal Component Analysis as a tool to explore star formation histories

Principal Component Analysis (PCA) is a well-known multivariate technique used to decorrelate a set of vectors. PCA has been extensively applied in the past to the classification of stellar and galaxy spectra. Here we apply PCA to the optical spectra of early-type galaxies, with the aim of extracting information about their star formation history. We consider two different data sets: 1) a reduced sample of 30 elliptical galaxies in Hickson compact groups and in the field, and 2) a large volume-limited (z<0.1) sample of ~7,000 galaxies from the Sloan Digital Sky Survey. Even though these data sets are very different, the homogeneity of the populations results in a very similar set of principal components. Furthermore, most of the information (in the sense of variance) is stored into the first few components in both samples. The first component (PC1) can be interpreted as an old population and carries over 99% of the variance. The second component (PC2) is related to young stars and we find a correlation with NUV flux from GALEX. Model fits consistently give younger ages for those galaxies with higher values of PC2.